Banner and Crosspost

Home    Overlay   Discord   Mutators   Maps   Integration   Links   About

November 26, 2019

Replay Analysis

Replay Analysis

While we are looking forward to Mengsk release, lets look back for a while. Over the weekend I've begun to learn some Python – inspired by Feardragon's introductory video. I expanded on that, and using sc2reader library, I went over my Co-op replays from last two years (3426 replays analyzed).

I will look at three things: Maps, Commanders and Masteries.


Only live Co-op games were analyzed, but this also includes mutations and various testing games (e.g. for Bug Hunts). The vast majority of games were played on Brutal difficulty (>99%). First let's look at game lengths. All following times are in the real-time and assuming faster game on speed on Brutal. In the following figure there is a decent number of very short games caused by quits, drops, wrong map selected or testing.

The average game length of a successful mission is 16.5 ± 3.7 minutes of real time.
(Converted into the game-time this would be 23.1 ± 5.18 minutes)

Now for other stats. Temple of the Past mission is the most played by wide margin. I'm not sure why, and it certainly doesn't feel like it. Perhaps some weekly mutations were more replayable on this mission – skewing the result.

Oblivion Express is a very good map, and I'm not surprised to see it in the second place. On the other side, Cradle of Death is the newest and relatively unpopular map, and so it makes sense to see it as last.

Unsurprisingly, maps with fixed lengths take more time.

Winrate data will be impacted by mutations and testing the most. My winrate on Dead of Night is the worst for the simple reason that I don't like the map and often quit at the start.


Following figures show how often did me or my ally pick each commander. Tychus, Zeratul and Stetmann were released during this period, and so they will be represented less. Also, my favorite commanders will be in theory seen less often in my allies. This would have to be adjusted for if you tried to get the true distribution.

We can see that Nova is very dominating. That's not too surprising, I even wrote a post about it (The popularity of Nova Terra). Both Zeratul and Tychus get pretty good representation despite being released later (more in the next figure).

Here are only games played after Stetmann release. But it's skewed by me playing a lot custom mutations with Ancalagon and Ziemson (playing a lot Raynor).

And finally my picks. This is over two years – Stetmann and Tychus would be a lot higher otherwise. Dehaka is one of my favorite commanders, and again I wrote a post about his design.

Now lets look how commander popularity changes over time. Tychus, Zeratul and Stetmann were released during this period. There isn't a big spike as you might have expected; this is because if I'm playing the commander, I won't get allies with the same commander. Stetmann seems to fall off more after his release compared to other commanders.

These figures are interesting, but given the limited data, they don't tell the story very accurately.


This section is probably the most interesting. Here we can see how popular is each mastery choice is. However, do note that we had several commander revamps, and some masteries did change or become more competitive.

Thank you for checking this out. It has been fun to learn a bit of Python and get some data from my replays. This is great for learning, but if you could pull any replay from Blizzard servers, you could show a lot more accurate and interesting data.

Recent posts



Powered by Blogger

Main post